sql: deal with retriable errors when using a new txn

[ajwerner]

In #46588 a bug was introduced when a retriable error was encountered while
using a new transaction for preparing. Prior to that commit, all error were
treated as not retriable. This was sort of a bummer. Retriable errors can
occur due to read within uncertainty. Before this PR, those retriable errors
would make their way to the client. Now we'll handle those retry errors
internally underneath connExecutor.prepare

Fixes #43251

Release note: None

[cockroach-teamcity]

This change is 

[andreimatei]

LGTM

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @ajwerner and @andreimatei)

---

pkg/sql/conn_executor_test.go, line 635 at r1 (raw file):

}

// This test ensures that when in an explicit transaction and statement

implicit. And maybe put it in the test's name too.

---

pkg/sql/conn_executor_test.go, line 659 at r1 (raw file):

	testDB.Exec(t, "CREATE TABLE foo (i INT PRIMARY KEY)")

	stmt, err := sqlDB.Prepare("SELECT * FROM [SHOW COLUMNS FROM foo]")

make this statement simpler now that we have the knob

[ajwerner]

TFTR!

bors r=andreimatei

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @andreimatei)

---

pkg/sql/conn_executor_test.go, line 635 at r1 (raw file):

Previously, andreimatei (Andrei Matei) wrote…

implicit. And maybe put it in the test's name too.

Done, this whole description was wrong.

---

pkg/sql/conn_executor_test.go, line 659 at r1 (raw file):

Previously, andreimatei (Andrei Matei) wrote…

make this statement simpler now that we have the knob

Done.

[craig]

Canceled (will resume)

[craig]

Build failed (retrying...)

• GitHub CI (Cockroach)

[ajwerner]

bors r+

[craig]

Build succeeded

• GitHub CI (Cockroach)